How to scrape Telegram members: what is possible and what is a myth

What you can and cannot extract

The line is drawn by the API, not by the tool. It is worth knowing exactly where it sits before buying anything.

You can get: the member list of a public group you have joined, including user IDs, usernames where set, display names, whether the account has Premium, last-seen bucket where the user allows it, and — for a chat you can read — who posted and when.

You cannot get: subscriber lists of channels. Channels are broadcast objects; their audience is not enumerable unless you are an administrator. Any tool advertising "channel subscriber scraping" is either scraping the linked discussion group, or scraping the people who commented — which is a different, smaller, and frankly more useful set.

You also cannot get: phone numbers of people who have not shared them with you, email addresses, or anything from private groups you are not a member of.

Why raw member counts are a bad target

A 200,000-member group sounds like a better source than a 3,000-member one. It usually is not.

Large public groups accumulate dead accounts, bots, and people who joined for a giveaway two years ago and never opened the app again. Messaging that list burns sessions on recipients who will never reply — and worse, a high proportion of unreachable or long-dormant accounts is exactly the pattern that gets outbound flagged.

The signals that actually predict a reply:

• Recent activity — posted or reacted in the last 30 days.
• Commented, not just joined — people who write in a discussion group are a different population from people who lurk.
• Profile completeness — an avatar and a bio correlate with a real, used account.
• Premium status — imperfect, but a paying user is a live user.

Three thousand recently-active commenters will out-convert two hundred thousand raw members, and cost a fraction of the accounts to work through.

Filtering that changes the outcome

Once you have a list, the filtering is where the campaign is won or lost. The filters worth applying, roughly in order of impact:

1. Last seen — drop anyone dormant beyond a month.
2. Bots and admins — obvious, routinely forgotten.
3. Keyword in bio — the cheapest intent signal available. Someone whose bio says "founder" or "trading" has told you what they are.
4. Language — the account's interface language is a better geography proxy than the group's nominal topic.
5. Deduplication across sources — the same people are in five groups in any niche. Messaging them five times is how you collect reports.

From list to campaign without wasting it

The sequence that avoids the usual failure: collect from several sources rather than one, deduplicate across all of them, filter for recent activity, then split the result across your account pool so each session works a slice at its own safe rate.

Do not export twenty thousand contacts and hand them to five accounts. That is four thousand each — an order of magnitude past what any session survives. The list size you can actually use is set by your pool, not by what you managed to collect.

In DarkTelegram the collection step feeds the campaign directly: audiences are gathered from groups and channels with the filters above, deduplicated, and pushed straight into a DM campaign or an inviter, distributed across accounts with per-account limits already applied. You can also export to CSV if the list is going somewhere else.